Multivariate analysis of variance
- The best choice is a set of DVs correlated with one another, but not overly so.
- When there is little correlation among DVs univariate F is acceptable.
- Unequal cell sizes and missing data are problematical for MANOVA.
- Low power can mean a non-significant multivariate effect but one or more significant Univariate F’s!
Extension of MANOVA, e.g., involving:
- Multiple DVs (e.g., Social, Campus, and Teaching/Education Satisfaction)
- Multiple Occasions (e.g., 1st, 2nd, 3rd year)
- Any number of IVs (e.g., Gender)
- Of interest are:
- Main effects of the multiple DVs
- Main effect of the multiple occasions
- Main effects of IVs
- Interactions b/w IVs and the DVs.